Structural role of Gly(193) in serine proteases: investigations of a G555E (GLY193 in chymotrypsin) mutant of blood coagulation factor XI.

In serine proteases, Gly(193) is highly conserved with few exceptions. A patient with inherited deficiency of the coagulation serine protease factor XI (FXI) was reported to be homozygous for a Gly(555) --> Glu substitution. Gly(555) in FXI corresponds to Gly(193) in chymotrypsin, which is the numbering system used subsequently. To investigate the abnormality in FXI(G193E), we expressed and purified recombinant FXIa(G193E), activated it to FXIa(G193E), and compared its activity to wild type-activated FXI (FXIa(WT)). FXIa(G193E) activated FIX with approximately 300-fold reduced k(cat) and similar K(m), and hydrolyzed synthetic substrate with approximately 10-fold reduced K(m) and modestly reduced k(cat). Binding of antithrombin and the amyloid beta-precursor protein Kunitz domain inhibitor (APPI) to FXIa(G193E) was impaired approximately 8000- and approximately 100000-fold, respectively. FXIa(G193E) inhibition by diisopropyl fluoro-phosphate was approximately 30-fold slower and affinity for p-aminobenzamidine (S1 site probe) was 6-fold weaker than for FXIa(WT). The rate of carbamylation of NH(2)-Ile(16), which forms a salt bridge with Asp(194) in active serine proteases, was 4-fold faster for FXIa(G193E). These data indicate that the unoccupied active site of FXIa(G193E) is incompletely formed, and the amide N of Glu(193) may not point toward the oxyanion hole. Inclusion of saturating amounts of p-aminobenzamidine resulted in comparable rates of carbamylation for FXIa(WT) and FXIa(G193E), suggesting that the occupied active site has near normal conformation. Thus, binding of small synthetic substrates or inhibitors provides sufficient energy to allow the amide N of Glu(193) to point correctly toward the oxyanion hole. Homology modeling also indicates that the inability of FXIa(G193E) to bind antithrombin/APPI or activate FIX is caused, in part, by impaired accessibility of the S2' site because of a steric clash with Glu(193). Such arguments will apply to other serine proteases with substitutions of Gly(193) with a non-glycine residue.

catalytic activity of all serine proteases (1,2). These residues are located at the entrance to the substrate binding pocket, and their geometry is stabilized by hydrogen bonds. Serine proteases hydrolyze peptide bonds via the formation of tetrahedral transition state intermediates. Stabilization of the transition state intermediate occurs through formation of hydrogen bonds between the oxyanion intermediate and the amido groups of residues Gly 193 and Ser 195 . The substrate binding sites in the enzyme involved in precise interactions are referred to as Sn, . . . S3, S2, S1, S1Ј, S2Ј, S3Ј, . . . SnЈ sites, and the amino acid residues of the substrate or inhibitor that occupy these sites are referred to as Pn, . . . P3, P2, P1, P1Ј, P2Ј, P3Ј, . . . PnЈ, respectively. These complementary sites permit specific alignment of the substrate/inhibitor with the catalytic triad and the oxyanion hole for enzymatic specificity (3). In trypsin-like serine proteases, Asp 189 is at the bottom of the primary S1 substrate binding site, and forms a salt bridge with the guanidino group of P1 Arg residues in the substrate/inhibitor.
Gly 193 , which is part of the oxyanion hole structural unit, is highly conserved in serine proteases with only a few exceptions (see "Discussion"). Several blood coagulation proteins with mutations at Gly 193 (chymotrypsin equivalent) in their protease domains have been reported. These include FXI 1 (Gly 555 3 Glu), FIX (Gly 363 3 Ala, Arg, Glu, or Val), and FVII (Gly 342 3 Glu or Arg) (4 -11). Each patient had normal plasma antigen levels associated with very low coagulant activity. In two cases (FXI G193E [555] 2 and FIX G193V [363]) the protein was activated normally (4,9). These data indicate that the functional abnormalities in these proteins stem from their inability to interact with their biological macromolecular substrate/inhibitors. This has been demonstrated for FIXa G193V (9).
FXI is a disulfide-linked homodimer with a molecular weight of ϳ160,000 (12). Deficiency of FXI results in a bleeding diathesis sometimes referred to as hemophilia C, and is most common in the Ashkenazi Jewish population (13,14). FXI can be activated to FXIa by FXIIa, thrombin, or by autoactivation (15,16). Upon cleavage of the Arg 15 -Ile 16 [Arg 369 -Ile 370 ] peptide bond, a heavy chain and a light chain are formed that are held 1 The abbreviations used are: FXI, factor XI; FIX, factor IX; FXI WT , wild-type factor XI; FXI G193E , factor XI with Gly 193 3 Glu mutation; FXIIa, factor XIIa; APPI, protease nexin-2/amyloid ␤ protein precursor Kunitz domain inhibitor; TBS, Tris-buffered saline; pNA, p-nitroaniline; S-2288, H-D-Ile-Pro-Arg-p-nitroanilide; S-2366, pyro-Glu-Pro-Arg-p-nitroanilide; pAB, p-aminobenzamidine; AT, antithrombin; BSA, bovine serum albumin; PEG, polyethylene glycol 8000; DFP, diisopropyl fluorophosphate; WT, wild type. 2 For comparison, the chymotrypsin amino acid numbering system is used throughout. Residue 555 in FXIa, 363 in FIXa, and 342 in FVIIa each correspond to residue 193 in chymotrypsin. Where necessary, the amino acid corresponding to a given protease is given in brackets (e.g. [555]). together by a disulfide bond (12,17). Thus, FXIa contains two heavy chains and two light chains. Each heavy chain contains four Apple domains, and each light chain a serine protease domain containing the catalytic triad His 57 [413], Asp 102 [464], and Ser 195 [557] (18). The new NH 2 terminus of the light chain contains the sequence Ile 16 -Val 17 -Gly 18 [370 -372]. The NH 2terminal Ile 16 [370], which is characteristic of serine proteases, inserts into the protease domain of FXIa, and its NH 2 group forms a salt bridge with the COOH group of Asp 194 [556]. This salt bridge is a defining feature of active serine protease formation (19). FXIa contributes to coagulation by activating FIX to FIXa in a Ca 2ϩ -dependent manner (20,21). The activation of FIX by FXIa has been shown to involve the Apple 2, Apple 3, and protease domains in FXIa and the activation peptide region and ␥-carboxyglutamic acid domains of FIX (22)(23)(24)(25).
In this report, we describe a series of experiments using physiologic macromolecular and small synthetic substrate/inhibitors to discern the nature of the proteolytic defect in a naturally occurring FXI mutant with a Glu substitution for Gly 193 [555]. The data on the rate of carbamylation of the NH 2 group of Ile 16 , which serves as an index of salt bridge formation with Asp 194 , as well as p-aminobenzamidine (pAB) binding and the rate of diisopropyl fluorophosphate (DFP) incorporation, strongly indicate that the S1 binding site and oxyanion hole are incompletely formed in the mutant enzyme. Binding of small synthetic substrates, however weak, provides sufficient energy to reorient the amido group of Glu 193 and restore the proper conformation of the oxyanion hole. Modeling efforts indicate that impairment of the interaction with physiologic macromolecular substrate/inhibitors is, in part, because of inaccessibility of the S2Ј site, which is attributable to a steric clash with Glu 193 [555].

EXPERIMENTAL PROCEDURES
Reagents-H-D-Ile-Pro-Arg-p-nitroanilide (S-2288) and pyro-Glu-Pro-Arg-p-nitoanilide (S-2366) were purchased from DiaPharma (West Chester, OH). Sodium boro[ 3 H]hydride was obtained from PerkinElmer LAS, Inc. Enhanced chemiluminescence (ECL) detection reagents were purchased from GE Healthcare. DFP was purchased from Calbiochem. Activated partial thromboplastin reagent was purchased from Beckman, and normal plasma was bought from George King (Overland Park, KS). Unfractionated heparin was purchased from Pharmacia Hepar, Inc. (Franklin, OH). Fatty acid-free bovine serum albumin (BSA), pAB, PEG 8000, Polybrene, and all other chemicals of the highest grade available were obtained from Sigma.
Proteins-Corn trypsin inhibitor, human FXIIa, antithrombin (AT), and human FIX were purchased from Enzyme Research Laboratories (South Bend, IN). The Kunitz domain of the protease nexin-2/amyloid ␤ protein precursor Kunitz protein inhibitor (APPI) was a gift from Dr. A. H. Schmaier (University of Michigan, Ann Arbor, MI).
SDS-Gel Electrophoresis-SDS-gel electrophoresis was done using the Laemmli buffer system (26). The acrylamide concentration used was 12% and the gels were stained with GelCode blue stain (Pierce).
Expression and Purification of Recombinant FXI Proteins-A Gly to Ala substitution was introduced into the human FXI cDNA (kindly provided by Dr. Dominic Chung, University of Washington, Seattle, WA) at base pair 1761 by site-directed mutagenesis. This substitution replaces the glycine residue normally found at amino acid position 555 (193 in chymotrypsin numbering) with glutamic acid. The cDNAs for wild type FXI and FXI G193E were ligated into mammalian expression vector pJVCMV, which contains a cytomegalovirus promoter, as previously described (24). 293 fetal kidney fibroblasts (50 ϫ 10 6 , ATCC CRL 1573) were cotransfected by electroporation (Electrocell Manipulator 600 BTX) with 40 g of pJVCMV-fXI-E555 construct and 2 g of plasmid RSVneo that contains a gene conferring neomycin resistance. Cells were grown in Dulbecco's modified Eagle's medium with 5% fetal bovine serum and 500 g/ml G418. G418 resistant clones were transferred to 96-well plates. Culture supernatants were tested for protein expression by enzymelinked immunosorbent assay using goat anti-human FXI polyclonal antibodies (Affinity Biologicals, Hamilton, Ontario, Canada). Expressing clones were expanded in 175-cm 2 flasks using Cellgro Complete media (Mediatech, Herndon, VA). Conditioned media was removed every 48 -72 h, supplemented with benzamidine (5 mM), and stored at Ϫ20°C pending purification.
Proteins were purified by monoclonal antibody affinity chromatography using anti-FXI IgG 1G5.1. Conditioned media (1 liter) was applied to the column, followed by washing with 25 mM Tris-HCl, pH 7.5, 100 mM NaCl (Tris/NaCl) with 5 mM benzamidine. Protein was eluted with Tris/NaCl containing 2.0 M sodium thiocyanate. Protein containing fractions were pooled and concentrated in an Amicon concentrator, dialyzed against Tris/NaCl, and stored at Ϫ80°C. Protein concentration was determined by the dye binding assay (Bio-Rad), and purity was assessed by SDS-PAGE. Purified proteins were homogenous on SDS-PAGE and had the correct molecular weight of a disulfide-linked homodimer. To prepare FXIa WT and FXIa G193E recombinant protein (ϳ300 g/ml) was incubated with 5 g/ml FXIIa at 37°C. Complete conversion of the zymogen to the heavy and light chains of FXIa was confirmed by reducing SDS-PAGE. The FXIIa in each activation reaction was inactivated by incubation with a 20-fold molar excess of corn trypsin inhibitor. The reactions were tested for residual FXIIa activity by monitoring the loss in activity following addition of corn trypsin inhibitor using a chromogenic assay.
[sialyl-3 H]Factor IX-3 H-Labeled FIX was prepared as described previously (28). The specific activity of the [sialyl-3 H]FIX was 2.1 ϫ 10 8 cpm/mg. The labeled preparation has 85% of the biological activity of the nonlabeled control as measured in an activated partial thromboplastin time assay (28). It showed a single band corresponding to FIX in both nonreduced and reduced SDS-PAGE (26).
Kinetics of Factor IX Activation-The rate of FIX activation was measured by quantifying the amount of radioactive peptide released at various times of incubation with FXIa. The procedure used was that described previously (28). Each reaction was carried out in TBS (50 mM Tris-HCl, pH 7.5, 150 mM NaCl) containing 0.5 mg/ml BSA and 5 mM Ca 2ϩ at 37°C. The concentration of FIX was varied from 2.0 to 25 g/ml. The concentrations of FXIa WT and FXIa G193E used were 20 ng/ml and 2.0 g/ml, respectively. Each reaction was started with the addition of FXIa, and at various times, 100-l aliquots were removed and added to 100 l of cold stopping buffer consisting of TBS, 50 mM benzamidine, 50 mM EDTA, and 5 mg/ml BSA. An equal volume (200 l) of 6% trichloroacetic acid was added and the reaction was centrifuged to precipitate the FIX/FIXa and FXIa proteins. One hundred-l aliquots of the supernatant were then removed and added to 4 ml of Aquasol 2 and counted for tritium in a Beckman LS 5000CE ␤-counter. Control experiments were performed in the absence of FXIa and were found to give a maximum of 1% of the total counts in the supernatant. These background counts were subtracted from each sample count prior to calculation of the initial FIXa activation rates. The amount of FIXa formed at a given time was obtained by averaging the results from three experiments and, the rates of activation were determined by least squares fitting of the initial data points to a straight line. Complete activation of FIX resulted in ϳ35% of the total counts in the trichloroacetic acid supernatant. Rates of activation were then plotted versus FIX concentration. The K m and k cat values were obtained using the Enzyme Kinetics program from Erithacus Software.
Measurement of S-2288 and S-2366 Amidolytic Activity of FXIa WT and FXIa G193E -Each reaction contained TBS with 5 mM Ca 2ϩ , 100 g/ml BSA, 0.5-1 nM FXIa WT , or 2 nM FXIa G193E (the results were normalized to 2 nM FXIa protein), and increasing amounts of either S-2288 or S-2366. The rate of release of pNA was measured using a Beckman DU800 spectrophotometer with a Kinetics module at 405 nm for 15 min. An extinction coefficient of 9.9 mM Ϫ1 cm Ϫ1 at 405 nM was used in calculating the amount of pNA released (29). The initial rate, which was linear, was then converted to micromolars substrate hydrolyzed per min. The program GraFit was used to determine the K m and V max using the Enzyme Kinetics Program from Erthicaus Software.
Inhibition of FXIa WT and FXIa G193E by AT in the Presence of Heparin-All reactions were carried out in 25 mM Tris-HCl, 100 mM NaCl, pH 7.4, plus 0.1% Tween 20 (Tris/NaCl/T) at 37°C. Recombinant FXIa (6 nM) was incubated with AT (30 nM to 3 M) in the presence of 1 unit/ml heparin. At various times, 20-l samples were removed and mixed with 80 l of Tris/NaCl/T containing 500 M S-2288 and 5 g/ml Polybrene (to dissociate the FXIa from heparin) in 96-well microtiter plates. The change in absorbance at 405 nm was measured in a Spec-traMax 190 microplate reader (Molecular Devices, Sunnyvale, CA), and the residual FXIa activity was determined at each time point. The residual activity was then plotted as a percent of initial activity and the first-order rate constants, k obs , for each concentration of AT used were obtained using the following equation, where A t and A 0 are the percent of FXIa activity at time t and 0 s, respectively. The values of k obs were then plotted against the AT concentration to obtain second-order rate constants.
Interaction of FXIa G193E and FXIa WT with APPI-These reactions were carried out using microtiter well plates from Dynatech. For experiments with FXIa WT , each reaction (150 l) contained 0.5 nM FXIa WT , 100 M S-2288, and increasing amounts of APPI in TBS with 5 mM CaCl 2 containing 100 g/ml BSA. For experiments with FXIa G193E , each reaction (150 l) contained 1 nM FXIa G193E , 500 M S-2288, and increasing amounts of APPI in TBS with 5 mM CaCl 2 containing 100 g/ml BSA. Absorbance at 405 nM (pNA release) was measured for up to 4 h in a Bio-Rad model 550 microtiter plate reader operated by Bio-Rad Labs microtiter manager PC software. The data were analyzed based upon the slow tight binding mechanism established for FXIa⅐APPI as follows, where FXIa⅐APPI is the steady state complex, which then isomerizes to XIa⅐APPI*. The forward and reverse rates for the initial binding are k 1 and k 2 , and k 3 and k 4 are the forward and reverse rates for the isomerization step. To calculate the binding parameters, the initial (v in ) and steady state (v st ) velocities were calculated using Equation 3, as described by Morrison and Walsh (30), where [P] represents the concentration of pNA formed at time t, v in and v st are, respectively, the rates of substrate hydrolysis before and after the steady state is achieved, and k obs is the rate of conversion of v in to v st . For these reactions, APPI was in vast excess to FXIa and the free APPI concentration did not change significantly during the course of the reaction. Substrate depletion also did not contribute to v st , as rates of control experiments done in the absence of APPI did not change significantly over time.
The values of k obs obtained were fitted to a single ligand binding site with a defined background using the following equation to obtain K d(app) , where the y intercept yields k 4 , the plateau value of k obs yields (k 3 ϩ k 4 ), and the midpoint of the curve yields K i(app) , which represents k 2 /k 1 (31). The value of K i (k 2 /k 1 ) was obtained from K i(app) using, where [S] is the S-2288 concentration. The K m values obtained with S-2288 were used to obtain K i . The overall K i , designated K i *, was derived using the following equation (30).
Determination of K dpAB of Binding of pAB to FXIa WT and FXIa G193E -The K d(app) of binding of pAB to FXIa G193E or FXIa WT was determined by its ability to competitively inhibit S-2288 hydrolysis. Details are provided in the legend to Fig. 2. The IC 50 (concentration of pAB required for 50% inhibition) was determined by fitting the data to the IC 50 four-parameter logistic equation of Halfman (32) given below.
Where y is the rate of pNA release in the presence of a given concentration of pAB represented by x, a is the maximum rate of pNA release in the absence of pAB, and s is the slope factor. Each point was weighted equally and the data were fitted to Equation 7 using the nonlinear regression analysis program GraFit from Erithcus Software. To obtain K d(pAB) values for the interaction of pAB with FXIa G193E and FXIa WT , we used the following equation described by Cheng and Prusoff (33) and further elaborated by Craig (34) where [S] is the S-2288 concentration. The K m values obtained for each protein using S-2288 (see Table I) were used to obtain K d(ppAB) . Inhibition of FXIa WT and FXIa G193E by DFP-Each reaction mixture contained 250 nM FXIa WT or FXIa G193E in TBS with 5 mM CaCl 2 and 100 g/ml BSA. Increasing amounts of DFP (2 M to 4 mM) were added to each reaction and incubated at room temperature for various times. At each time point, 5-l aliquots of each reaction were removed and added to 155 l of TBS/BSA containing 625 M S-2288. The change in absorbance at 405 nm was measured in a SpectraMax 190 microplate reader (Molecular Devices) and the residual FXIa activity was determined at each time point. The residual activity was then plotted as a percent of initial activity and the first-order rate constants, k obs , for each concentration of DFP used were obtained using Equation 1, where A t and A 0 are the percent FXIa activity at time t and 0 s, respectively. The values of k obs were then plotted against the DFP concentration to obtain second-order rate constants. Carbamylation of Ile 16 in FXIa G193E and FXIa WT by Reaction with NaNCO-These experiments were performed as described by Camire (34). Briefly, each reaction mixture contained 1 M FXIa mutant or normal protein in 20 mM Hepes, 0.15 M NaCl, 0.1% PEG 8000, 2 mM CaCl 2 , pH 7.5 (HBSP). Each experiment was performed in the absence and presence of pAB (10ϫ the K D pAB) and each reaction was started by the addition of 0.2 M NaNCO. The final pH after the addition of NaNCO was 7.5. Every 30 min, a 5-l aliquot was removed and added to 145 l of HBSP containing 500 M S-2288. The residual activity was determined from the initial linear rates of hydrolysis using a Beckman DU 800 spectrophotometer. The residual activity was plotted as a percent of initial activity and k obs for carbamylation were determined using Equation 1.
Molecular Modeling-The three-dimensional structure information of the zymogen and activated serine protease domains of FXI WT and FXI G193E as well as the complexes of APPI and the serine protease domain of FXIa were derived using software from Biosym/MSI (San Diego, CA) and the Swiss-Model server using the optimize mode (36 -38 G193E with APPI. Bulk solvent is excluded from the protease-inhibitor complex, thus, it is anticipated that hydrogen bonds and ionic interactions can be accurately evaluated and play an important role in specificity. The relative positions of the inhibitor and protease domains were maintained, and adjustments were only made to the side chains. Hydrophobic/van der Waals, hydrogen bonds, and ionic interactions were observed between each protease-inhibitor complex. These interactions were taken into consideration in evaluating each protease-inhibitor complex, and it was assumed that all potential hydrogen bond donors and acceptors would participate in these interactions.

FXIa WT and FXIa G193E Amidolytic Activity toward Small
Synthetic Substrates-We initially studied the effects of the Glu 193 substitution on protease catalytic activity using the tripeptide chromogenic substrates S-2366 and S-2288. These data are presented in Table I. The catalytic efficiency of FXIa G193E was ϳ18.5-fold lower (ϳ12-fold increase in K m and ϳ1.5-fold decrease in k cat ) for S-2366 and ϳ11.5-fold lower (ϳ7-fold increase in K m and ϳ1.7-fold decrease in k cat ) for S-2288. These data indicate that the active site is impaired in FXIa G193E .
FIX Activation by FXIa WT and FXIa G193E -During physiologic coagulation, the macromolecular substrate for FXIa is FIX. We examined activation of FIX by FXI WT and FXI G193E . These data are presented in Fig. 1. The values for K m were similar for the two proteases (FXIa WT , 145 nM and FXIa G193E , 170 nM); however, k cat was ϳ300-fold slower for activation by FXIa G193E . The K m data are consistent with previous observations indicating that exosites distant from the FXIa protease domain, such as the Apple 2 and Apple 3 domains, are important in binding to FIX (22)(23)(24)(25). One should note that the k cat value in this situation is influenced by the interaction of the activation peptide cleavage sites of FIX with the active site of FXIa, as well as by catalysis of the peptide bond by FXIa. Thus, the ϳ300-fold difference in k cat for WT and mutant FXIa represents a cumulative effect of restricted binding and catalysis at the active site. The ϳ15-fold average reduction in catalytic efficiency for synthetic substrates (Table I) versus ϳ300-fold reduction for FIX (Fig. 1) could be accounted for by misalignment at the S2Ј/P2Ј binding sites in FIX and FXIa G193E (see "Discussion").
FXIa G193E and FXIa WT Inhibition by AT and APPI-AT and APPI are two known physiologic inhibitors of FXIa. AT belongs to the serpin family of inhibitors and APPI to the Kunitz family of inhibitors (similar to bovine pancreatic trypsin inhibitor). In contrast to small synthetic substrates and the activation peptide cleavage sites in FIX, AT and APPI are likely to make extensive contacts with the protease domain of FXIa. For this reason, we hypothesized that binding of these inhibitors to FXIa G193V may be severely impaired in comparison to synthetic substrate hydrolysis and FIX activation. The second-order rate constant, k, for binding of AT to FXIa WT was 6.1 M Ϫ1 s Ϫ1 and to FXIa G193V was 8 ϫ 10 Ϫ4 M Ϫ1 s Ϫ1 . These data indicate that binding of AT to FXIa G193V is ϳ8,000-fold weaker than binding to FXIa WT .
Binding data for APPI inhibition of FXIa WT and FXIa G193E are presented in Table II. The initial rapid equilibrium binding (K i ) of APPI to the mutant enzyme was impaired ϳ6000-fold. Furthermore, the isomerization step (k 3 ) that leads to the tight binding complex was also impaired ϳ20-fold. However, dissociation of the tightly bound complex was equivalent for both enzymes. Thus, from the AT and APPI binding data, one may conclude that FXIa G193V may not be locked into an active enzyme conformation. Instead it may exist in an equilibrium state that fluctuates between the active and zymogen forms of the enzyme. The experiments described in the following sections were designed to test this concept.
Binding of pAB to FXIa G193E and FXIa WT -The inhibitor pAB is known to bind to the S1 site of serine proteases, and was employed to investigate the integrity of the S1 site in FXIa G193E . The data are presented in Fig. 2. Notably, FXIa G193E bound to pAB with 6-fold weaker affinity (K i ϭ 126 M versus 21 M) than to FXIa WT . Thus, the S1 site is impaired in FXIa G193E , in agreement with the synthetic substrate hydrolysis data presented in Table I.
Inhibition of FXIa G193E and FXIa WT by DFP-One of the underlying features of serine proteases is the presence of the oxyanion hole, which develops upon conversion of the zymogen to the enzyme form. DFP specifically reacts with Ser 195 and contains an oxyanion that enables it to be used as a probe to test the integrity of the oxyanion hole (48). As shown in Fig. 3, FXIa G193E inhibition by DFP was ϳ30-fold slower when compared with inhibition of FXIa WT (20 M Ϫ1 min Ϫ1 versus 610 M Ϫ1 min Ϫ1 , respectively). This demonstrates that the oxyanion hole in the mutant enzyme is not properly formed, and that the amide N of Glu 193 [555] in FXIa G193E is not pointing precisely toward the oxyanion hole.
Carbamylation of Ile 16 Using NaNCO-The data in the previous two sections indicate that the S1 site and the oxyanion hole in the mutant enzyme are not properly formed. The development of the S1 site and oxyanion hole in serine proteases requires formation of a salt bridge between the amino group of Ile 16 (35). We therefore examined the effect of pAB of the S1 site on the stability of the salt bridge between Ile 16 [370] and Asp 194 [556] in FXIa G193E . Notably, pAB occupancy of the S1 site in FXIa G193E led to an ϳ8-fold decrease in the rate of carbamylation (Table III), indicating substantial stabilization of the salt bridge. Interestingly, occupancy of the S1 site of FXIa WT by pAB also resulted in an ϳ2-fold slower rate of carbamylation (Table III), suggesting that pAB further stabilizes the wild type enzyme in a more active conformation. Remarkably, the presence of pAB corrected the rate of carbamylation of FXIa G193E to that of FXIa WT . Thus, the presence of pAB in the S1 site appears to largely correct the abnormality in the main chain conformation involving residues 189 -194 [551-556] in the mutant enzyme. DISCUSSION We have investigated the structural role of Gly 193 [555] in the proper conformation and activity of the mutant coagulation serine protease FXIa G193E . The substitution of Glu for Gly 193 [555] was based upon a naturally occurring mutation found in a patient with a history of excessive bleeding (4). In preliminary work reported in abstract form (4), zymogen FXI G193E could be cleaved normally between Arg 15 [369] and Ile 16 [370] to generate the protease FXIa G193E . Thus, bleeding associated with FXI G193E is likely because of the inability of FXIa G193E to function as an enzyme. FXIa was chosen for the present study because it has significant amidolytic activity and a well characterized physiologic macromolecular substrate/inhibitor profile.
Replacement of Gly 193 [555] by Glu in FXIa may have two consequences: 1) it may change the conformation around the latent active site in the zymogen, and 2) it may not allow the specific conformational changes that accompany conversion of the zymogen to the enzyme to occur. Analysis of a model for the human FXI protease domain suggests that Glu at position 193 [555] can be easily accommodated in the zymogen form of the protein (Fig. 4A)   , which line one side of the S1 binding pocket are also shown. C, model of FXIa G193E in complex with APPI. The protease domain backbone of FXIa G193E is shown in magenta and that of APPI in cyan. Helices are depicted as cylinders, ␤-sheets as long thick arrows, and turns in the loops with short thin arrows. The NH 2 and COOH termini are labeled N and C, respectively. The interacting residues in FXIa G193E and APPI are colored by atom type. Residues from FXIa are labeled in white and residues from residues. The Glu 193 [555] side chain would then point into a cavity that is filled with solvent in most serine proteases. The Glu 193 side chain may also be able to make a hydrogen bond with the hydroxyl group of Tyr 228 [590]. Thus, it is expected that Glu substitution at position 193 [555] can be easily accommodated in FXI without significant structural consequences.
In contrast, in the active form of the enzyme, the polypeptide backbone at Gly 193 [555] assumes a conformation compatible only with a Gly residue. Gly 193 [555] is located in position 3 of a type II hairpin loop, and hasconformation parameters of (105, Ϫ10) in the modeled FXIa WT structure. These parameters are compatible with those reported in crystal structures of many serine proteases, and places this residue in a region of the Ramachandran plot that is compatible only with a Gly residue. Replacement of Gly 193 [555] with any other residue will introduce a ␤-carbon that will have a steric conflict with the carbonyl O of residue 192 [554] (Fig. 4B). Relief of this conflict can only be achieved by a change in the conformation around the 192-193 [554 -555] peptide bond. One possible way to achieve this is to flip the bond, thereby converting the type II turn into a type I turn. In this process, rotation () around the N-C␣ iϩ1 bond and rotation () around the C␣-CЈ i changes the orientation of the peptide bond without affecting side chain positions or the main chain conformation (49). The maximum activation barrier for a concerted flip of this type is Ͻ3 kcal/mol (49). In the case of FXIa G193E , this results inconformation parameters of (Ϫ61, Ϫ45) in the modeled structure (Fig. 4B). Support for the premise that the 192-193 [554 -555] bond can be readily flipped in FXIa G193E comes from structures of the serine protease factor VIIa (bound to tissue factor) in complex with a certain class of inhibitor (50). Thus, it would appear that the facile interconversion of the 192-193 peptide bond occurs with a relatively low energy barrier. In both the type II and type I turns, the Glu 193 [555] side chain points into the solvent and may make a hydrogen bond with Arg 39 [395] (Fig. 4B).
The experimental evidence supports the argument that the conformation of the 192-193 [554 -555] bond is altered when Glu occupies position 193 [555] instead of Gly. An altered conformation in the mutant would be expected to perturb the structure of the S1 binding site, oxyanion hole, and salt bridge formation between Ile 16 [370] and Asp 194 [556]. The data clearly demonstrate that the S1 site (Fig. 2), oxyanion hole (Fig.  3), and Ile 16 -Asp 194 salt bridge formation (Table III) Table III show that occupancy of the S1 site alone can correct the impairment in formation of the salt bridge in FXIa G193E . Similarly, binding of DFP (Fig. 3) should also reorient the amide N of residue 193 [555] into the proper conformation for formation of a hydrogen bond with its oxyanion. The observations that both pAB and DFP binding are impaired in FXIa G193E is strong evidence that the S1 site and oxyanion hole are not preformed in the mutant enzyme, but can be reordered upon binding of a substrate that occupies the S1 site and/or oxyanion hole.
Two crystal structures of serine proteases with bound inhibitors support our argument that occupancy by the substrate/ inhibitor can correct the defects in the FXIa G193E S1 site and oxyanion hole. These are human brain trypsin with Arg at position 193 and the S1 site inhibitor (Ref. 51, Protein Data Bank code 1h4w) and Trimeresurus stejnejeri plasminogen activator with Phe at position 193 and the chloromethylketone inhibitor (Ref. 52, Protein Data Bank code 1bqy). Furthermore, a flip of the 192-193 peptide bond upon substrate binding is also supported by crystal structure and biochemical data for Staphylococcus aureus exfoliative toxins A (53-55) and B (55). The crystal structure of mouse glandular kallikrein-13 with Asp at position 193 and no inhibitor is also known (Ref. 56, Protein Data Bank code 1ao5). Some residues lining both sides of the S1 site have very high B factors, suggesting that this region is somewhat mobile in the absence of inhibitor occupancy. In this structure, although the amide N of residue 193 is pointing toward the oxyanion hole to some extent, this region appears to be quite mobile. Thus, the S1 site may not be completely formed in the absence of occupancy of the S1 site.
The hydrolysis of the macromolecular substrate, FIX, is significantly more impaired than is hydrolysis of synthetic substrates by FXIa G193E . This could be because of the side chain of Glu 193 occupying part of the S2Ј site where the P2Ј residue of FIX is expected to reside (see below), in addition to the abnormal conformation of the S1 site and oxyanion hole. The sequence in FIX at its NH 2 -terminal most cleavage site (Arg 145 -Ala 146 ) from P2 to P2Ј is Thr-Arg-Ala-Glu, and at the COOHterminal cleavage site is (Arg 180 -Val 181 ) Thr-Arg-Val-Val. Considering the P2Ј residues, it appears that the steric clash would be larger at the Arg 145 -Ala 146 cleavage site for two reasons: 1) the Glu at this site is larger than Val, and 2) Glu at the P2Ј position would produce a charge repulsion. Thus, we expect that the Arg 145 -Ala 146 peptide bond in FIX would be cleaved very slowly in comparison to cleavage of the Arg 180 -Ala 181 peptide bond by mutant FXIa G193E .
The inhibition of FXIa G193E by AT and APPI was even more impaired than was activation of FIX. This may be due primarily to the larger contact regions between the protease domain of FXIa and the two inhibitors when compared with the limited contact region with the FIX activation peptide. A model of APPI in complex with FXIa G193E is shown in Fig. 4C. The electrostatic and hydrophobic interactions are described in the figure legend. The indicated interactions are in agreement with data presented in abstract form by Navaneetham et al. (27). Clearly, APPI are labeled in cyan. As in A and B, the protease domain numbering is based upon chymotrypsin and the APPI numbering is based upon that of bovine pancreatic trypsin inhibitor. Specifically, Tyr 60  the presence of Glu at position 193 [555] will change the electrostatic potential around the active site and interfere with the binary collision with AT or APPI. Moreover, Glu 193 [555] will have a steric clash with the P2Ј Met residue of APPI as shown in Fig. 4C. Data presented in Table II indicate that the isomerization step that leads to the formation of the tight complex after the initial binary collision with FXIa G193E is also severely affected. This is expected because minor rearrangements of Glu 193 [555] in FXIa G193E and the P2Ј residue Met 17 in APPI must take place for a stable complex to form. However, once a stable complex is formed, dissociation of the complex does not appear to be affected.

CONCLUSIONS
The Gly 193 [555] residue of FXIa, a serine protease involved in blood coagulation, was changed to Glu to study the role of Gly 193 in serine proteases. The experimental and modeling data presented in this paper indicate that the S1 site, oxyanion hole, and salt bridge formation between Ile 16 [370] and Asp 194 [556] are impaired in the mutant enzyme. However, occupancy of the active site by substrate/inhibitors can correct these defects. The hydrolysis of macromolecular substrates and binding of macromolecular inhibitors can be further impaired by a steric clash between the S2Ј site of the enzyme and P2Ј residue of the substrate/inhibitor. Enzymes with residues other than glycine at position 193, such as human brain trypsin (Arg 193 ) (51), T. stejnejeri plasminogen activator (Phe 193 ) (52), and mouse glandular kallikrein-13 (Asp 193 ) (56) would be compatible with substrate/inhibitors that have small side chains at the P2Ј position. This will introduce a new level of structural plasticity and selectivity in serine proteases. Thus, it would appear that enzymes with Gly at position 193 will be more active but not have selectivity at the S2Ј site; whereas, enzymes with non-Gly residues at position 193 will be an order of magnitude less active but have a restricted S2Ј site and selectivity for their substrate/inhibitors. Bode and co-workers (52) have previously elaborated such plasticity and substrate/inhibitor selectivity in this class of proteases.